Glossary

The terms, explained briefly.

Plain definitions of the ideas behind sovereign AI, each with a short explanation and links to go deeper.

AI voice agent: An AI voice agent is a system that holds spoken conversations, typically on phone lines, using speech recognition, a language model and speech synthesis, and can complete tasks in back-end systems. More
Embeddings: Embeddings are numerical vectors that represent the meaning of text, so that passages with similar meaning have similar vectors. More
Fine-tuning: Fine-tuning is further training of a pre-trained model on an organisation's own examples so that it learns specific tasks, formats or vocabulary. More
Human-in-the-loop: Human-in-the-loop is a design where AI prepares or recommends an action and a named person approves it before it takes effect, with the decision logged. More
Model Context Protocol (MCP): The Model Context Protocol (MCP) is an open protocol that lets AI applications such as Claude, ChatGPT or Cursor connect to external tools and data sources through a standard interface. More
Model gateway: A model gateway is a single routing layer through which every AI model call passes, making it possible to swap providers, control access and track usage and cost. More
On-premise AI: On-premise AI is AI software, including models, indexes and logs, that runs on an organisation's own servers rather than as a third-party cloud service. More
Reranking: Reranking is a second step in AI search where a specialised model reorders retrieved passages by how well they answer the question, improving accuracy. More
Retrieval-augmented generation (RAG): Retrieval-augmented generation (RAG) is a technique where an AI system first retrieves relevant passages from a knowledge base and then has a language model answer using those passages, usually citing them. More
Small language model (SLM): A small language model (SLM) is a compact language model, typically with a few billion parameters or fewer, that runs on modest hardware and can be fine-tuned to perform well on narrow tasks. More
Sovereign AI: Sovereign AI is AI that an organisation runs on infrastructure it controls, with models it can inspect or replace and data, logs and decisions that stay under its own jurisdiction. More
Vector database: A vector database stores numerical representations of content, called embeddings, and finds the items most similar in meaning to a query. More

Start with two weeks of evidence, not a sales call.

A fixed-price discovery sprint, credited against whatever comes next. Or write to sales@deepvox.ai.

Book a discovery sprint